South Korea Opens First Humanoid Robot Data Factory To Train Robots On Real Home Tasks

Korea humanoid data factory robot

South Korea opened its first humanoid robot data factory on November 3. Aju Press reported the opening on November 3. The facility will train robots on real home tasks at national scale, while KIMM and ETRI share control of the site. The Korea humanoid data factory brings together 12 humanoids on about 1,000 square meters over two floors.

At the opening, remote controlled humanoids demonstrated everyday chores, in a display reminiscent of a robot duck walking on its own. First, one robot opened a fridge. Another handled laundry. A third pressed buttons and levers on a sofa and a car. The demos showed the factory’s core idea. Robots learn by doing real tasks under human guidance, and every movement therefore becomes training data.

That data feeds vision language action models. These AI systems let a humanoid see a scene, understand an instruction, and, at the same time, move its body. The factory records each task in detail and uses the records to improve the models, because better records mean better robots. Over time, the loop should produce robots that handle new homes and new objects with less human help.

KIMM president Ryu Seog-hyeon said large scale, high quality task data is a matter of national competitiveness. The Korea humanoid data factory answers that need with a shared source of data researchers control. South Korea wants to reduce its dependence on foreign data and foreign models for humanoid development.

Korea humanoid data factory laundry robot
A researcher operates a humanoid near a washing machine. (Photo: KIMM)

The site will expand from 12 humanoids to 25. It will also open in stages to outside companies, universities, and research institutes from late 2027. That staged opening matters, particularly for smaller teams. Smaller teams cannot build a data factory alone, because shared access spreads the benefit across the local industry.

The factory sits inside K-Moonshot, a national program with 12 missions that include humanoid robots. The goal is a world class national platform within three years. From 2027, the Next-X Humanoid Challenge will pick a final humanoid design through competition. As a result, the factory’s data will feed that contest.

Funding backs the ambition. The Korea humanoid data factory is part of that funded plan. The Ministry of Planning and Budget set aside 2.3 trillion won for humanoid development through 2030. That sum equals about 1.7 billion dollars. The plan counts 1,080 domestic humanoids already in use, and public events such as a Hangzhou marathon robot show how visible the field has become. It targets an 80 percent local share for core parts, up from 45 percent today.

The Korea humanoid data factory also connects to KAIROS, KIMM’s own humanoid line. The first KAIROS version is due in April 2027. Data from the factory should help that timeline. A robot trained on thousands of recorded tasks needs less trial and error, because its training already covers the task.

The first tasks focus on home chores the demos already showed. Fridge handling, laundry sorting, and sofa and car controls give the models everyday examples. Each example adds to a library other robots can learn from, and the library therefore grows with every session.

Korea humanoid data factory KIMM humanoid
A humanoid built by KIMM meets a visitor. (Photo: Asian Robotics Review)

That library is the Korea humanoid data factory’s real product, more than any single robot. Each recorded session also notes failures, such as a missed grip or a wrong button press. Those failure notes guide the next round of training, because robots learn as much from mistakes as from clean runs.

Humanoids have already entered public events across Asia, from a robot duck to a Hangzhou marathon robot and a robot race up Mount Tai. The factory takes that public excitement into systematic work. The regional race in embodied AI is moving fast. South Korea is betting that owned data, not only owned hardware, decides who leads.

The Korea humanoid data factory will face a practical test in its first year. Collecting data is one task, and turning it into reliable robots is another. Homes are messy, objects vary, and a model trained in one kitchen must still cope with the next. The factory’s two floors give researchers room to test that jump.

South Korea’s bet is clear: control the data, and the robots follow. The factory opened with 12 humanoids and a national plan behind it. A working data loop will give the country both the robots and the training records, progress of the kind shown by a robot race on Mount Tai. The Korea humanoid data factory makes that loop possible.

Stay Connected with The MES Times

Get the latest tech, AI and gadget news and updates wherever you prefer.

Follow on Google Discover

Add The MES Times to Preferred Sources and see more of our stories in Google Search and Top Stories.

AI & Emerging Tech

Aizaz Khan is a technology journalist at The MES Times, leading coverage on artificial intelligence, humanoid robotics, and silicon hardware. Holding a degree in AI and Robotics where he graduated top of his class he approaches tech reporting with an engineer's eye rather than PR talking points. His work focuses on dissecting research papers, neural model architectures, and physical robotics systems to explain their real-world impact and limitations.

0 comments Oldest comments first